Re^3: Tracing memory leak

That example code leaks an SV. You return a newly-minted RV that points to a HV. In the $test assignment, the value of that RV is copied to $test; the original RV leaks.

In your production code, try examining the value of the perl internals variable PL_sv_count (e.g. write a little C wrapper function that returns its value to perl space). This var shows the total number of SVs allocated. If it keeps growing, it means that you're leaking SVs.

Dave.

Comment on Re^3: Tracing memory leak

Replies are listed 'Best First'.
Re^4: Tracing memory leak by halfcountplus (Hermit) on Sep 14, 2011 at 20:43 UTC
That example code leaks an SV. No, it does not. As mentioned, running this within a while(1) loop, even for several minutes (ie, millions and millions of iterations) produces 0 growth. That is unequivocally no leak. `my $count = 0; while (1) { my $test = eg(); print $count++."\n"; reportHash($test); reportScalar($test->{x}); $test = undef; }` [download] Try this yourself. If even a single byte were leaking per iteration, this would be pretty obvious pretty quickly in a monitor. I believe the problem you perceive is circumvented by the use of _noinc -- the value returned does not increment the reference count of the SV. Otherwise, it would be impossible to return an RV without a leak. I will look at PL_sv_count tho.	[reply] [d/l]
Re^5: Tracing memory leak by dave_the_m (Monsignor) on Sep 14, 2011 at 22:11 UTC
No, it does not Ah, you're right. XS does an implicit sv_2mortal() of the returned RV, which is why it doesn't get leaked. I guess Inline::C must do something similar (I've never used Inline::C). The problem I perceived however wouldn't have been circumvented by the _noinc usage. Dave.	[reply] [d/l]
Re^6: Tracing memory leak by halfcountplus (Hermit) on Sep 14, 2011 at 23:04 UTC
The problem I perceived however wouldn't have been circumvented by the _noinc usage. Heh-heh -- I was going to add something like "or else you have assumed something more chimeric about the RV" since based on the empirical it seemed to me this is just a pointer value but your explication explicates (namely the implicit sv_2mortal, and the implication that the RV is a scalar tracked by perl, not just a pointer value), cool. I am very new to perl internals. PL_sv_count does indicate an issue but I am too beat for this evening. Thanks!	[reply]